Caching the stable hash of Ty within itself
Instead of computing stable hashes on types as needed, we compute it during interning.
This way we can, when a hash is requested, just hash that hash, which is significantly faster than traversing the type itself.
We only do this for incremental for now, as incremental is the only frequent user of stable hashing.
As a next step we can try out
* moving the hash and TypeFlags to Interner, so projections and regions get the same benefit (tho regions are not nested, so maybe that's not a good idea? Would be nice for dedup tho)
* start comparing types via their stable hash instead of their address?
The `PrettyPrinter` changes formatting of array size and integer
constants based on `-Zverbose`, so its implementation cannot be used in
legacy symbol mangling.
Fix ICE when passing block to while-loop condition
We were incorrectly delaying a bug when we passed _any_ block (that evaluated to `()`) to a while loop. This PR makes the check a bit more sophisticated.
We should only suppress the error here in cases that are equivalent to those we find in #93574 (i.e. only while loop conditions that have destructuring assignment expressions in them).
Fixes#93997
cc `@estebank` who added this code
I would not be opposed to removing the delay-bug altogether, and just emitting this error always. I much prefer duplicate errors over no errors.
* Recover from invalid `'label: ` before block.
* Make suggestion to enclose statements in a block multipart.
* Point at `match`, `while`, `loop` and `unsafe` keywords when failing
to parse their expression.
* Do not suggest `{ ; }`.
* Do not suggest `|` when very unlikely to be what was wanted (in `let`
statements).
Only create a single expansion for each inline integration.
The inlining integrator used to create one expansion for each span from the callee body.
This PR reverses the logic to create a single expansion for the whole call,
which is more consistent with how macro expansions work for macros.
This should remove the large memory regression in #91743.
Apply noundef metadata to loads of types that do not permit raw init
This matches the noundef attributes we apply on arguments/return types.
Fixes (partially) #74378.
This was necessary before, because `declare_raw_fn` would always apply
the default optimization attributes to every declared function,
and then `attributes::from_fn_attrs` would have to remove the default
attributes in the case of, e.g. `#[optimize(speed)]` in a `-Os` build.
However, every relevant callsite of `declare_raw_fn` (i.e. where we
actually generate code for the function, and not e.g. a call to an
intrinsic, where optimization attributes don't [?] matter)
calls `from_fn_attrs`, so we can simply remove the attribute setting
from `declare_raw_fn`, and rely on `from_fn_attrs` to apply the correct
attributes all at once.
Rollup of 7 pull requests
Successful merges:
- #94396 (1 - Make more use of `let_chains`)
- #94397 (Document that pre-expansion lint passes are softly deprecated)
- #94399 (Add test for #79465 to prevent regression)
- #94409 (avoid rebuilding bootstrap when PATH changes)
- #94415 (Use the first codegen backend in the config.toml as default)
- #94417 (Fix duplicated impl links)
- #94420 (3 - Make more use of `let_chains`)
Failed merges:
r? `@ghost`
`@rustbot` modify labels: rollup
Apply noundef attribute to all scalar types which do not permit raw init
Beyond `&`/`&mut`/`Box`, this covers `char`, enum discriminants, `NonZero*`, etc.
All such types currently cause a Miri error if left uninitialized,
and an `invalid_value` lint in cases like `mem::uninitialized::<char>()`.
Note that this _does not_ change whether or not it is UB for `u64` (or
other integer types with no invalid values) to be undef.
Fixes (partially) #74378.
r? `@ghost` (blocked on #94127)
`@rustbot` label S-blocked
Use the first codegen backend in the config.toml as default
It is currently hard coded to llvm if enabled and cranelift otherwise.
This made some sense when cranelift was the only alternative codegen
backend. Since the introduction of the gcc backend this doesn't make
much sense anymore. Before this PR bootstrapping rustc using a backend
other than llvm or cranelift required changing the source of
rustc_interface. With this PR it becomes a matter of putting the right
backend as first enabled backend in config.toml.
cc ```@antoyo```
Document that pre-expansion lint passes are softly deprecated
The pre-expansion lint pass has been softly deprecated since https://github.com/rust-lang/rust/pull/69838. Every once in a while I see someone mention it as a possibility, only get the feedback that it's deprecated. This PR officially documents that the method is soft deprecated to have a single point of truth for it.
That's it. Have a great rest of the day 🙃
---
* See [rust#69838](https://github.com/rust-lang/rust/pull/69838)
* See [rust-clippy#5518](https://github.com/rust-lang/rust-clippy/pull/5518)
LLVM really dislikes this and will assert, saying something along the
lines of:
```
rustc: llvm/lib/MC/MCSubtargetInfo.cpp:60: void ApplyFeatureFlag(
llvm::FeatureBitset&, llvm::StringRef, llvm::ArrayRef<llvm::SubtargetFeatureKV>
): Assertion
`SubtargetFeatures::hasFlag(Feature) && "Feature flags should start with '+' or '-'"`
failed.
```
This expands the existing `#[must_use]` check in `unused_attributes`
to lint against pretty much everything `#[must_use]` doesn't support.
Fixes#93906.
Avoid query cache sharding code in single-threaded mode
In non-parallel compilers, this is just adding needless overhead at compilation time (since there is only one shard statically anyway). This amounts to roughly ~10 seconds reduction in bootstrap time, with overall neutral (some wins, some losses) performance results.
Parallel compiler performance should be largely unaffected by this PR; sharding is kept there.
It is currently hard coded to llvm if enabled and cranelift otherwise.
This made some sense when cranelift was the only alternative codegen
backend. Since the introduction of the gcc backend this doesn't make
much sense anymore. Before this PR bootstrapping rustc using a backend
other than llvm or cranelift required changing the source of
rustc_interface. With this PR it becomes a matter of putting the right
backend as first enabled backend in config.toml.
rustc_trait_selection: adopt let else in more places
Continuation of #89933, #91018, #91481, #93046, #93590, #94011.
I have extended my clippy lint to also recognize tuple passing and match statements. The diff caused by fixing it is way above 1 thousand lines. Thus, I split it up into multiple pull requests to make reviewing easier. This PR handles rustc_trait_selection.
No branch protection metadata unless enabled
Even if we emit metadata disabling branch protection, this metadata may
conflict with other modules (e.g. during LTO) that have different branch
protection metadata set.
This is an unstable flag and feature, so ideally the flag not being
specified should act as if the feature wasn't implemented in the first
place.
Additionally this PR also ensures we emit an error if
`-Zbranch-protection` is set on targets other than the supported
aarch64. For now the error is being output from codegen, but ideally it
should be moved to earlier in the pipeline before stabilization.
Beyond `&`/`&mut`/`Box`, this covers `char`, discriminants, `NonZero*`, etc.
All such types currently cause a Miri error if left uninitialized,
and an `invalid_value` lint in cases like `mem::uninitialized::<char>()`
Note that this _does not_ change whether or not it is UB for `u64` (or
other integer types with no invalid values) to be undef.
Restrict query recursion in `needs_significant_drop`
Overly aggressive use of the query system to improve caching lead to query cycles and consequently ICEs. This patch fixes this by restricting the use of the query system as a cache to those cases where it is definitely correct.
Closes#92725 .
This is essentially a revert of #90845 for the significant drop case only. The general `needs_drop` still does the same thing. The hope is that this is enough to preserve the performance improvements of that PR while fixing the ICE. Should get a perf run to verify that this is the case.
cc `@cjgillot`
